Composite Zoned Namespace Cloud Storage Architecture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The adoption of NVMe-SSD based All-Flash-Arrays in cloud storage is hindered by the unpredictability of garbage collection processes, leading to unpredictable performance, and limitations in Zoned NameSpace (ZNS) SSDs such as a fixed number of active zones and potential performance reduction from random writes, which increases errors as more SSDs are included.
Innovation Solution
Implementing a Composite Zoned Name-Space (C-ZNS) architecture within cloud storage devices, which includes a housing with blades containing storage devices and processors that manage data writing by initializing buffers, determining command parameters, and writing data only when the command is error-free, thereby avoiding garbage collection during sequential writes and handling zone limits and random writes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If ZNS SSD is used to avoid garbage collection and provide predictable performance, then performance predictability is improved, but the number of active zones is limited to 14 and zone size is fixed
Solution Approach 1:
The patent divides the storage system into multiple independent blades, where each blade contains its own ZNS SSDs and can manage zones independently. This segmentation allows the system to overcome the 14-zone limit of individual SSDs by distributing zones across multiple blades, while each blade maintains the performance predictability benefits of ZNS architecture.
Solution Approach 2:
The patent introduces a new dimension of scalability by moving from a single SSD with limited zones to a multi-blade architecture where zones can be expanded across multiple physical devices. This dimensional change allows the system to scale the number of active zones beyond the 14-zone limitation while maintaining ZNS performance characteristics.
2Productivity
If ZNS SSD with fixed zone size (e.g., 1 GB per zone) is used, then sequential write performance is optimized, but host application flexibility is reduced when requesting larger write operations
Solution Approach 1:
The patent merges multiple fixed-size zones across different blades to provide the host application with effectively larger write capabilities. By combining the zone resources of multiple blades, the system allows host applications to perform writes larger than the individual zone size without requiring the application to close and reopen zones, thus maintaining both sequential performance and application flexibility.
Solution Approach 2:
The multi-blade architecture provides universal access to zone resources, allowing the system to handle both small sequential writes (utilizing individual zones for optimal performance) and large writes (spanning multiple zones across blades) through a unified interface, eliminating the need for application-specific optimizations.
3Quantity of substance
If multiple SSDs are included in the storage device, then storage capacity is increased, but the number of errors increases due to limitations in each SSD
Solution Approach 1:
The patent extracts the error-prone elements (fixed zone limitations and random write vulnerabilities) from individual SSDs by implementing a blade architecture where each blade operates as an isolated unit with its own error handling. This extraction allows errors to be contained within individual blades rather than propagating system-wide, enabling the addition of more storage devices without proportionally increasing the overall error rate.
Data Source
AI summary
Systems, methods, and non-transitory computer-readable media for providing a cloud storage device implementing C-ZNS architecture. The cloud storage device including a housing and a plurality of blades, with at least one blade including a plurality of storage devices and an electronic processor. The electronic processor is configured to receive data and a command from a host application to write the data to a corresponding zone of a first storage device of the plurality of storage devices, initialize a buffer in response to receiving the data, store the data in the buffer that is initialized, determine one or more parameters in response to receiving the command, determine that the command is error-free based on the one or more parameters, and write the data from the buffer into the corresponding zone in response to determining that the command is error-free.